Feb

6

Google DeepMind’s new AI system can solve complex geometry problems. Its performance matches the smartest high school mathematicians and is much stronger than the previous state-of-the-art system.

DeepMind says it tested AlphaGeometry on 30 geometry problems at the same level of difficulty found at the International Mathematical Olympiad, a competition for top high school mathematics students. It completed 25 within the time limit. The previous state-of-the-art system, developed by the Chinese mathematician Wen-Tsün Wu in 1978, completed only 10.

A collection of Olympiad geometry problems.

H. Humbert comments:

But it can't solve talent retention.

Google DeepMind scientists in talks to leave and form AI startup

Humbert H. writes:

If anyone were to ask, perhaps the real hidden value of the system could be for example in the application in discovery of new materials. Of course, the biggest question and the current AI can't solve immediately is how to syntheses the new materials in real experiment timely and to verify and validate the properties. If it could break new grounds in materials, one of the fantastic 7 is also on it too.

H. Humbert responds:

Depends on whether materials synthesis can be described via some set of rules. I don't know enough about it to see it one way or the other. I expect new drug discovery by pharma companies which is now being transitioned to digital molecule exploration from lab based experimentation to eventually use AI, at the every least for new protein synthesis which has both chemical and spatial folding problems and is a less general problem than "materials". DeepMind seems different from large language models that have been in the news lately in that it operates on much less data and is generally used to find better solutions to problems that are similar to games and have a more contained set of "rules" as opposed to mimicking human intelligence by mapping how humans answer questions after analyzing huge data sets.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search